Randomization Interview Q&A — Deep Answers

Overview

Constraint-pattern drills test whether you can write constraints; this topic tests whether you can explain them. Interviewers grade the explanation questions harder than the coding ones, because a memorized snippet survives one follow-up and a real mental model survives five. Every answer here is structured the way a senior engineer delivers it: the direct answer in one or two sentences, a supporting example, the follow-up the interviewer is already planning, and the contrast between the junior phrasing and the senior phrasing of the same fact.

Read these after the drills. Where a question leans on a drill (sum width traps, implication skew), the answer states the principle and points back rather than re-deriving it.

Sub-topics

1. Q&A: rand, randc & randomize() — the fundamentals every screen starts with.

2. Q&A: dist, inside, soft, solve-before — the operator-semantics round.

3. Q&A: Controlling Randomization — rand_mode, constraint_mode, inline, overrides.

4. Q&A: Solver Semantics — bidirectionality, seeds, randc interaction, skew, state vars.

5. Q&A: Arrays & Nested Objects — sizes, allocation, sums, 2D randomization.

6. Q&A: Methodology & Debug — reproduction, constraint placement, error injection, coverage link.

Legend: [QA]

  HOW THE RANDOMIZATION INTERVIEW ESCALATES            [QA]

  Screen      "rand vs randc?"  "what does randomize() return?"
     │
     ▼
  Operators   "dist := vs :/"   "inside vs dist"   "soft ordering?"
     │
     ▼
  Control     "freeze one field"  "relax a class constraint inline?"
     │
     ▼
  Solver      "is solving bidirectional?"  "why did adding an
     │         implication skew my distribution?"
     ▼
  Design      "where do constraints belong — txn, sequence, test?"
              "how do you reproduce a random failure?"

  Each level assumes fluent answers at the level below.
  Senior offers the follow-up before it is asked.

Key takeaways

• Answer pattern: direct answer first, example second, trade-off or follow-up third.

• Junior recites syntax; senior states semantics and volunteers the consequence.

• The solver model (simultaneous, bidirectional, uniform over solutions) underlies every question.